Image extraction system for extracting patterns such as characters, graphics and symbols from image having frame formed by straight line portions

Abstract

In an image extraction system, an extracting part for extracting wide lines, an extracting part for extracting narrow lines and a frame detector detect a frame from a pattern which is extracted by a connected pattern extracting part. An attribute adder adds attributes of a character (graphic and symbol inclusive), frame, and a contact pattern of the character and frame to a partial pattern, and a separating part separates the frame from the contact pattern. An intersection calculator calculates intersections of the character and frame, and the calculated intersections are associated by an intersection associating part. An interpolator obtains a character region within the frame and interpolates this region based on the associated intersections. A connection confirming part confirms a connection of the pattern with respect to the extracted character pattern, and patterns confirmed of their connection are integrated in a connected pattern integrating part to thereby extract the character.

What is claimed is: 
     
       1. An image extraction system for extracting patterns including at least one of characters, graphics and symbols from an image having a frame formed by straight line portions said image extraction system comprising: connected pattern extracting means for extracting partial patterns having connected pixels from the image;   straight line extracting means for extracting straight line portions from the partial patterns;   attribute adding means for categorizing the partial patterns into a first pattern including only portions of the pattern which do not touch the frame and portions of the frame which do not touch the pattern, and a second pattern including portions of the pattern which touch the straight line portion extracted by said straight line extracting means;   separating means for separating the straight line portions from the second pattern;   intersection calculating means for calculating intersections of the patterns and the straight line portions from the second pattern;   intersection associating means for associating the intersections which are calculated by said intersection calculating means;   interpolating means for interpolating a region of the pattern within the straight line portions forming the frame based on the intersections which are associated by said intersection associated means; and   connection confirming means for confirming a connection of a pattern with respect to the extracted partial patterns of the pattern, said connection confirming means extracting a confirmed connection of the pattern and feeding back the portions of the pattern which cannot be confirmed to said intersection associating means, said intersection associating means re-associating the intersections calculated by said intersection calculating means with respect to the portion of the pattern fed back by said connection confirming means, said interpolating means interpolating a region of the pattern within the straight line portions forming the frame based on the intersections which are re-associated by said intersection associating means.   
     
     
       2. The image extraction system as claimed in claim 1, wherein said connection confirming means obtains a connection component by labeling only candidates of the pattern including portions of the pattern which touch the frame. 
     
     
       3. The image extraction system as claimed in claim 1, wherein said connection confirming means limits re-associating the intersections in said intersection associating means to where a straight line connecting the intersections exists on the straight line portion. 
     
     
       4. The image extraction system as claimed in claim 1, wherein said intersection associating means comprises: means for extracting both end points of a line segment which forms the pattern as intersection candidates to be associated from intersections where said line segment and the straight line portions connect;   means for calculating a number of intersection candidates at each of two contours respectively located on inner and outer sides of and forming each straight line portion and calculating a distance between each pair of associated candidates by associating the candidates when the calculated number of candidates at the two contours match; and   means for confirming the association by checking each distance between the candidates based on a line width of the straight line portion.   
     
     
       5. The image extraction system as claimed in claim 1, wherein said intersection associating means comprises: first means for extracting both end points of a line segment which forms the pattern as intersection candidates to be associated out of intersections where said line segment and the straight line portions connect;   second means for calculating a number of intersection candidates at each of two contours respectively located on inner and outer sides of and forming each straight line portion;   third means for calculating a middle point between two intersections of each line segment forming the pattern when the calculated number of candidates at the two contours do not match, and for calculating a distance between the middle points which are calculated for the two contours on the inner and outer sides forming the straight line portion;   fourth means for associating the middle points and the intersection candidates forming the middle points when the calculated distance between the middle point is less than or equal to a predetermined threshold value which is determined by a line width of the straight line portion; and   fifth means for obtaining one of a slope of a line segment forming the pattern at the intersection candidates and a straight line perpendicularly intersecting the straight line portion, with respect to the innersection candidates which cannot be associated by said fourth means, and for associating the intersection candidates depending on a continuity of said line segment based on the slope of said line segment and the intersection candidate by including the intersection candidates which are already associated.   
     
     
       6. The image extraction system as claimed in claim 1, wherein said intersection associating means approximates a pattern which is formed by the association by a rectangle using labeling, calculates a ratio of a size of the rectangle and one of a size of the pattern and a size of the frame, and cancels association of the intersections when the ratio exceeds a predetermined threshold value. 
     
     
       7. The image extraction system as claimed in claim 1, wherein said interpolating means comprises: means for obtaining straight lines connecting the intersections which are associated by said intersection associating means and for identifying a portion inside of said straight lines as a region of the pattern;   means for obtaining a slope of a contour line of a line segment forming the pattern at an intersection of a line segment that is not associated, and for obtaining a cross-point of two straight lines which pass through each intersection and have the obtained slope of the contour;   means for identifying a range surrounded by said two straight lines and an ouster contour line of the frame, including the outer contour line of the frame, as a region of the pattern when the cross-point is located within the frame; and   means for identifying a range within the frame between said two straight lines and the intersection of said two straight lines and the outer contour line of the frame, excluding the outer contour line of the frame as a region of the pattern when the cross-point is located outside the frame.   
     
     
       8. The image extraction system as claimed in claim 1, wherein said interpolating means interpolates a region within the straight line portions using straight lines corresponding to a line width of the pattern, so that a continuity of the pattern is restored. 
     
     
       9. The image extraction system as claimed in claim 1, which further comprises: connected pattern integrating means for determining whether a portion of the partial patterns exists inside the frame based on a position of the extracted frame, integrating the partial patterns as a constituent element of the pattern if a decision result is yes, determining whether a ratio of a size of the integrated pattern and a size of the frame is within a range of a predetermined threshold value, employing the integrated pattern if the ratio is within the range, and extracting each original partial pattern without integration if the ratio is outside the range.   
     
     
       10. The image extraction system as claimed in claim 1, wherein said intersection associating means calculates a distance between candidates of the intersections at two contours on inner and outer sides forming each straight line portion with respect to each of line segments forming the pattern and are associated by a one-to-one association, and calculates an average line width of the line segments forming the pattern from an average value of the calculated distances. 
     
     
       11. The image extraction system as claimed in claim 1, wherein said separating means obtains connection components using labeling with respect to a separated pattern which is obtained by separating the straight line portions from an original pattern including the straight line portions and the pattern, and eliminates noise in the straight line portions based on an area of each connection component. 
     
     
       12. The image extraction system as claimed in claim 1, wherein said intersection associating means comprises: means for associating the intersections based on a distance between each pair of character line segments which form one character and are included in the first pattern and a continuity of a slope of each of the character line segments; and   means for adaptively changing references used to determine the distance and continuity depending on a line width of the character line segments.   
     
     
       13. The image extraction system as claimed in claim 1, wherein said intersection associating means comprises: means for associating intersections to maintain continuity according to a priority sequence set based on at least the distances among the intersections having different labels using the intersections calculated by said intersection calculating means, with respect to one of the pattern and a portion of the pattern a connection of which cannot be confirmed by said connection confirming means.   
     
     
       14. The image extraction system as claimed in claim 13, wherein said priority sequence is a table of conditions for associating the intersections. 
     
     
       15. The image extraction system as claimed in claim 1, wherein said intersection associating means comprises: means for associating intersections to maintain continuity according to a predetermined priority sequence using the intersections calculated by said intersection calculating means, with respect to a pattern or a portion of the pattern a connection of which cannot be confirmed by said connection confirming means.   
     
     
       16. The image extraction system as claimed in claim 15, wherein said predetermined priority sequence is set in a table of conditions for associating the intersections based on a kind or type of character such as alphabets including Japanese Katakana characters and Japanese Hiragana characters. 
     
     
       17. The image extraction system as claimed in claim 1, wherein said straight line extracting means comprises: first means for extracting a straight line portion from each partial pattern out of the partial patterns extracted by said connected pattern extracting means, with respect to a block frame which is partitioned into frames of 1 character each; and   second means for determining whether the straight line portion extracted by said first means forms a part of the frame.   
     
     
       18. The image extraction system as claimed in claim 17, wherein said second means determines whether the straight line portion is a part of the frame depending on whether the straight line portion extracted by said first means is perpendicular to and reaches a straight line portion on an outermost part within the partial pattern.